Available online at www.sciencedirect.com
Progress in
Oceanography
Progress in Oceanography 75 (2007) 179–199
www.elsevier.com/locate/pocean
Persistently declining oxygen levels in the interior waters of
the eastern subarctic Pacific
Frank A. Whitney *, Howard J. Freeland, Marie Robert
Institute of Ocean Sciences, Fisheries and Oceans Canada, P.O. Box 6000, Sidney, BC, Canada V8L 4B2
Available online 17 August 2007
Abstract
Fifty years of measurements at Ocean Station Papa (OSP, 50N, 145W) show trends in the interior waters of the subarctic
Pacific that are both impacted by short term (few years to bi-decadal) atmospheric or ocean circulation oscillations and by
persistent climate trends. Between 1956 and 2006, waters below the ocean mixed layer to a depth of at least 1000 m have been
warming and losing oxygen. On density surfaces found in the depth range 100–400 m (rh = 26.3–27.0), the ocean is warming
at 0.005–0.012 C y1, whereas oxygen is declining at 0.39–0.70 lmol kg1 y1 or at an integrated rate of 123 mmol m2 y1
(decrease of 22% over 50 years). During this time, the hypoxic boundary (defined as 60 lmol O2 kg1) has shoaled from 400
to 300 m. In the Alaska Gyre, the 26.2 isopycnal occasionally ventilates, whereas at OSP 26.0rh has not been seen at the ocean
surface since 1971 as the upper ocean continues to stratify. To interpret the 50 year record at OSP, the isopycnal transport
of oxygenated waters within the interior of the subarctic Pacific is assessed by using a slightly modified ‘‘NO’’ parameter
[Broecker, W., 1974. ‘‘NO’’ a conservative water-mass tracer. Earth and Planetary Science Letters 23, 100–107]. The highest
nitrate–oxygen signature in interior waters of the North Pacific is found in the Bering Sea Gyre, Western Subarctic Gyre and
East Kamchatka Current region as a consequence of winter mixing to the 26.6 isopycnal. By mixing with low NO waters
found in the subtropics and Okhotsk Sea, this signature is diluted as waters flow eastward across the Pacific. Evidence of low
NO waters flowing north from California is seen along the coasts of British Columbia and SE Alaska. Oxygen in the subsurface waters of the Alaskan Gyre was supplied 60% by subarctic and 40% by subtropical waters during WOCE surveys,
whereas such estimates are shown to periodically vary by 20% at OSP. Other features discernable in the OSP data include
periods of increased ventilation of deeper isopycnals on an 18 year cycle and strong, short term (few month) variability
caused by passing mesoscale eddies. The potential impacts of declining oxygen on coastal ecosystems are discussed.
Crown Copyright 2007 Published by Elsevier Ltd. All rights reserved.
Keywords: Dissolved oxygen; Oxygen consumption; Hypoxia; Nitrates; Water temperature; Ocean stratification; Subarctic Pacific Ocean;
Alaska Gyre; Ocean Station P
1. Introduction
Animal life in our oceans requires oxygen. Various organisms adapt to a wide range of oxic conditions,
some requiring the saturated conditions of the surface ocean to support their rapid metabolism, while others
*
Corresponding author. Tel.: +1 250 363 6346.
E-mail address: [email protected] (F.A. Whitney).
0079-6611/$ - see front matter Crown Copyright 2007 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.pocean.2007.08.007
180
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
tolerate brief periods of anoxia (Davis, 1975; Gray et al., 2002). A range of effects occur as oxygen levels
decline, progressing from slowed growth rates, to metabolic impairments, to death. Since oxygen is continuously consumed in seawater via the remineralization of organic matter, the ocean depends on continuous
inputs to the intermediate and deep ocean from the atmosphere. However, the subarctic Pacific is a strongly
stratified ocean realm because of the freshness (low salinity) of the surface layer. Over much of this area, the
ocean is mixed to about 100–125 m during winter storms (Freeland et al., 1997; Watanabe et al., 2001). Only in
the mixed layer are oxygen levels found near 100% saturation. Below this, oxygen levels rapidly decline, reaching saturations of less than 10% below 500 m (Aydin et al., 2004).
As the atmosphere of our planet warms, so do its oceans. This is evident in the eastern subarctic Pacific
(Freeland et al., 1997) where both near shore and open ocean data records show ocean warming rates similar
to those being measured in the atmosphere. Possibly, the poleward transport of freshwater from the subtropics
is also stabilizing the surface layer of the subarctic Pacific (Wong et al., 2001; Freeland et al., 1997), as has
been observed in the North Atlantic (Curry et al., 2003). As a result, the upper ocean is becoming more buoyant and stratification is strengthening. In surveys conducted over the past half century, oxygen declines are
being observed in all areas of the subarctic Pacific. A 30 year record in the Oyashio Current off the north coast
of Japan (Fig. 1) shows declining and oscillating trends in waters between 190 and 630 m, on density surfaces
from 26.7 to 27.2 (Ono et al., 2001). Rates of oxygen decline varied between 0.6 and 1.3 lmol kg1 y1 over
the period 1968–1999. Concomitant with this trend was a tendency for the density of the upper ocean to
decline and these waters to warm at about 0.007 C y1.
Other regions of the western Pacific see similar trends. A 50 year decline of 0.4–0.8 lmol O2 kg1 y1 was
found on the 26.9–27.0 isopycnal surfaces in the Western Subarctic Gyre and Sea of Okhotsk (Andreev and
Kusakabe, 2001). Also in the adjoining East/Japan Sea, observed oxygen decreases of between 0.5 and
2.0 lmol kg1 y1 were observed in deep waters between 1932 and 2000 (Watanabe et al., 2003; Kang
et al., 2004).
World Ocean Circulation Experiment (WOCE) surveys have been repeated in most regions of the subarctic
Pacific (Fig. 2). One of the persistent trends coming out of this work is the striking rate of oxygen decline over
about a decade. Watanabe et al. (2001) used WOCE sections along 47N (1985 and 1999) and 165E (1987 and
2000) to calculate apparent oxygen utilization (AOU, the estimated amount of oxygen consumed since waters
were last in equilibrium with the atmosphere) increases of 6 lmol kg1 y1. In general, they found oxygen levels decreased by 18 mol m2 over 13 years and concluded that the formation rate of intermediate ocean waters
had substantially declined over a 15 year period, perhaps longer based on results of Ono et al. (2001). These
66
Russia
Alaska
62
Bering Sea
Latitude (N)
58
Okhotsk
Sea
54
50
BSG
Z9
EKC
AG
AS
BC
P
OSIW
AC
WSG
P4
WA
46
NPC
CC
OR
42
38
140E
Subarctic Boundary
OC
CUC
KC
160E
180
160W
140W
CA
120W
Longitude
Fig. 1. Map of the North Pacific Ocean showing major currents (EKC, East Kamchatka Current; OSIW, Okhotsk Sea Intermediate
Water; OC, Oyashio Current; KC, Kuroshio Current; NPC, North Pacific Current; CC, California Current; CUC, California
Undercurrent; AC, Alaska Current; AS, Alaska Stream) and gyres (BSG, Bering Sea gyre, WSG, Western Subarctic gyre; AG, Alaska
gyre). Also identified are Ocean Station P (P) at 50N, 145W, station P4 at 48.66N, 126.67W, British Columbia (BC), Washington
(WA), Oregon (OR) and California (CA).
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
181
Fig. 2. Temperature anomaly in C (monthly anomalies computed against 1956–1991 averages) and seawater density as sigma theta
(monthly averages) at Ocean Station Papa from 1956 to 2005. Two mesoscale eddies are evident in: (1) 1960 and (2) 1974, these waters
being characterized by depressed isopycnal surfaces, elevated temperature and low oxygen. After 1981, sampling was too infrequent to
reliably resolve eddies.
findings were expanded by adding data from an eastern Pacific WOCE repeat section (Emerson et al., 2001;
Emerson et al., 2004). These authors reported oxygen decreases of up to 60–80 lmol kg1 over the 13–14 year
period between surveys, but averaging in the 10–40 lmol kg1 range for waters in the 33–34 salinity range
(densities of 26.5–27.0). Kumamoto et al. (2004) reported very similar findings along WOCE section P17
through the Alaska Gyre. Again, oxygen declined by as much as 60 lmol kg1 between the early 1990s and
2001 in the density range 26.2–27.0 (200–600 m). Feely et al. (2004) used CFCs to age waters then calculated
oxygen utilization rates (OUR) for various ocean basins. For the subarctic Pacific, they estimate an OUR of
4.1 lmol kg1 y1 for the depth range 200–900 m.
The subarctic Pacific is an ocean basin bordered by eastern Asia and western North America (Fig. 1), and
by a distinct subarctic boundary that separates the warm, saline waters of the subtropics from the cool, fresh
waters of the subarctic (Dodimead et al., 1963). Within the subarctic region, ocean circulation is generally
cyclonic, with the eastward flowing North Pacific Current (NPC) bifurcating into subtropical (California Current, CC) and subarctic (Alaska Current, AC) branches as it approaches shore. The AC and buoyant coastal
currents feed into the Alaskan Stream (AS) north of the Alaskan Gyre, with some portion of this water flowing southward around the western side of the Alaskan Gyre (AG). Flow through the Bering Sea is not shown
but results in a strong southward flowing East Kamchatka Current (EKC), which is modified by outflow of
Sea of Okhotsk Intermediate Water (SOIW) into the Oyashio Current (OC). The warm Kuroshio Current
(KC) and the OC create the NPC. The subarctic boundary is imbedded within the NPC, separating subarctic
OC waters from subtropical KC waters. Three cyclonic gyres form in the subarctic Pacific and Bering Sea, the
Western Subarctic Gyre (WSG), the Alaskan Gyre (AG) and the Bering Sea Gyre (BSG).
Only in the western subarctic Pacific can isopycnal surfaces in the 26.5–26.7 range exchange gases with the
atmosphere. The 26.6 isopycnal is seen outcropping within the domain of the East Kamchatka and Oyashio
Currents (Ono et al., 2001; Mecking et al., 2006). From this region, winter ventilated waters submerge into the
interior of the Pacific. You (2005) followed the path of the North Pacific Intermediate Waters (NPIW), centred
on the 26.8 isopycnal surface, eastward from the coast of Japan. He suggests mixing of Okhotsk Intermediate
Water and Gulf of Alaska Intermediate Water along the North Pacific Current produces NPIW waters with
the appropriate CFC age. Mixing of various waters results in a product with increased density (cabbeling),
with the result that these waters sink into the subtropics. Ueno and Yasuda (2003) show that waters on the
182
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
26.5–26.8 isopycnal surface flow almost directly eastward, taking 5–7 years to travel from the edge of the eastern subarctic ventilation region (155E) to Ocean Station Papa (OSP) at 50N, 145W. Their model estimates
a 15 year transit time on the 27.2 isopycnal.
Waters in the North Pacific Current acquire heat and salt from the subtropics as they cross the ocean (Mecking et al., 2006). These waters contain more oxygen than those within the Alaska Gyre, and are the oxygen
source for the interior waters of the eastern subarctic Pacific. Between WOCE sections near 180 and 145W,
integrated oxygen utilization rates were calculated to be 2.1 ± 0.4 M m2 y1 for waters found between the
26.5 and 27.2 isopycnals (the 132–706 m range; Aydin et al., 2004).
Much of what has recently been discovered about the circulation of intermediate waters in the subarctic
Pacific is based on WOCE data sets. Almost all of these open ocean results are dependent on repeat surveys,
separated by periods of 10–15 years. Detail is provided for the ventilation site off the coast of Japan (Ono
et al., 2001) otherwise not much open ocean time-series data is available except at OSP. Tabata (1989)
reviewed data from this site for the period 1956–1983 and concluded declines in oxygen levels were mostly
not significant. The trend on the 26.8 isopycnal surface was for oxygen to decrease by 0.7 lmol kg1 y1.
We now have 50 years of data at OSP and revisit this data set to discuss patterns of oxygen decline in the
subarctic Pacific. Our analysis of oxygen trends begins with a review of the OSP data set, next we define areas
where interior waters of the subarctic Pacific exchange gases with the atmosphere, then use a conservative
property of seawater that is based on a combination of oxygen and nitrate (Broecker, 1974) to trace oxygen
transport into the NE subarctic Pacific, and finally suggest reasons for oxygen decline and impacts this trend
may have on biota.
2. Methods and data sources
Three major data sets are used to describe oxygen distribution and transport in the subarctic Pacific; the
WOCE (World Ocean Circulation Experiment) one time surveys (http://whpo.ucsd.edu/about.htm), the
OSP time series data (http://www.pac.dfo-mpo.gc.ca/sci/osap/projects/linepdata/default_e.htm) and Argo
(http://www.argo.net/). A brief description of data reliability for each source follows.
WOCE data go through quality control steps which compare survey sections with both historical data and
with other WOCE sections that occupy the same stations. For routine parameters such as temperature, salinity
and oxygen, these data are of excellent quality. For example, the Institute of Ocean Sciences contributed
WOCE data from sections P15 and P1W, the latter in collaboration with Pacific Oceanological Institute of
Vladivostok, Russia. During IOS surveys, data accuracy was 0.002 C for temperature, 0.003 for salinity
and 1.02 lmol kg1 for oxygen (Whitney and Barwell-Clarke, 1997). A repeat survey of WOCE line P1
(Fukasawa et al., 2004) suggested that temperature differences as low as 0.0007 C might be detectable.
Argo temperature and salinity data are collected from profiling autonomous floats, typically from 2000 m
to the surface every 10 days. These data then go through quality control procedures set up by each country
contributing to the program. Data are compared to historical values or recent CTD casts throughout the
ocean to identify floats that may have drifting sensors. Generally, salinity data are accurate to better than
0.01. Temperature sensors are believed to have relatively slow drift from factory calibrations over time, therefore errors are thought to be considerably less than 0.01 C. When these data are used to calculate density, the
error will be less than 0.01. Since we use Argo data only to show areas of winter outcropping of the 26.5 isopycnal surface, this potential error is trivial.
The data record from OSP consists of an era from 1956 to 1981 when Weatherships occupied this station
almost continually and the period from 1981 to 2006 during which research vessels transited there 2–5 times
annually. During the Weathership era, casts for temperature, salinity and oxygen were taken as frequently as
several times per week.
Sampling and analytical procedures for samples collected along Line P (a survey line running from southern British Columbia to Ocean Station Papa) and at OSP have evolved over the 50 years of this program.
Early sampling relied on unprotected reversing thermometers to provide depth measurements (estimated to
be accurate to ±15 m at 4000 m depth), Loran C for position (navigators could be 10 nautical miles off station), reversing thermometers for temperature (accurate to ±0.02 C) and older style water samplers for salinity and oxygen. OSP temperature and salinity data were carefully edited to remove bad data from leaking
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
183
samplers, misfires, poor sample handling and analyses (data sets reviewed by Tabata and Peart, 1985; Tabata
and Brown, 1994). During the WOCE era, IOS improved its sampling procedures by switching from hydro
casts (sampling bottles on wire) to rosette casts. Depth accuracy of pressure sensors on various CTDs (conductivity–temperature–depth probes) improved to better than 2 m at full ocean depth, temperature was generally accurate to 0.002 C, and salinity analyses were accurate to 0.002 with the use of IAPSO standard
seawater and Guildline Autosal salinometers. Oxygen procedures have been consistent throughout the 50 year
period with all samples being analyzed by the Winkler method (Carpenter, 1965). Still, a large number of analysts (perhaps >100) carried out this work. Therefore, this data set was hand edited to remove obvious outliers.
These were primarily identified as single anomalous points in a profile. Nutrient methods are those described
by Barwell-Clarke and Whitney (1996) and rely on modified Technicon procedures. Nitrate is accurate to
±0.2 lmol kg1. Since 1987 nutrients have been analyzed at sea. Prior to this, samples were frozen and results
contain uncertainty large enough to make data unreliable for the analysis of ventilation trends.
To track changes in temperature over 50 years, monthly anomalies have been computed. Steps taken to
produce a monthly climatology for OSP include: (1) for months when more than one cast was made, data were
averaged so that years with many casts in a month do not bias the climatology; (2) all data for various months
between 1956 and 1991 were used to calculate monthly averages for this period; (3) monthly averages were
then used to derive monthly anomalies. Only 13 stations were samples along Line P up to 1981, whereas
27 stations were sampled following the Weathership era.
Data for the three isopycnal surfaces of interest included all data within certain ranges (26.45–26.55, 26.65–
26.75 and 26.88–26.92). When data were scarce in a geographic region or within a time period, interpolation of
values to the desired density surface was carried out by assuming linear change between data points above and
below the density surface in a profile. Potential densities (rh) are implied throughout this paper whenever we
refer to isopycnal surfaces. Differences between temperature and potential temperature to depths of 600 m are
on the order of 0.04 C, creating only a minor difference (0.003) between density and potential density (rt
and rh).
When discussing ocean circulation, oceanographers commonly use oxygen changes in units not easily
applied to biological studies. The following conversions represent a level commonly used to define the hypoxic
threshold (Gray et al., 2002):
2:0 mg l1 ¼ 1:40 ml l1 60 lmol kg1 :
In the 3–5 C temperature and between salinities of 33–34, Apparent Oxygen Utilization (AOU) levels for
hypoxia would be 6.0 ± 0.2 ml l1 (260 lmol kg1) which is equivalent to an oxygen saturation of 23%.
3. Results
3.1. The 50 year record at Ocean Station P
Ocean Station P, at 50N, 145W, is located between the southern edge of the Alaska Gyre (AG) and the
North Pacific Current (NPC) within the subarctic waters of the NE Pacific (Fig. 1). It is the terminal station of
a 1400 km long survey line (Line P) that starts on the south coast of British Columbia. P4, a Line P station at
which oxygen and nutrients have been routinely measured since 1987, sits on the continental slope in 1300 m
of water. Because Weatherships occupied OSP for most of the period from 1956 to 1981 and research cruises
have continued to sample a few times per year since, an extensive temperature, salinity and oxygen data set has
been collected with which to observe impacts of climate on the subarctic ocean.
Temperature anomalies for OSP (Fig. 2) highlight some of the processes creating variability in the oxygen
record. In general, the entire upper water column was cool in the decade from 1962 to 1972, with waters being
>2 C below the 1956–1991 average for much of this period. Since 1972, the T anomaly has been generally
positive for the upper 500 m. Cool events of several months are associated with La Niñas in late 1983,
1989 and 1999. Maximum subsurface temperatures, occasionally exceeding the climatology by >3 C, were
recorded in association with El Niños of 1958, 1983, 1992, 1994–1995 and 1998, also with mesoscale eddies
that passed through OSP in 1960 and 1974. These eddies, containing warm, low oxygen core waters of coastal
origin, are characterized by strongly depressed isopycnal surfaces (Whitney and Robert, 2002).
184
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
Density surfaces in intermediate waters (Fig. 2) fluctuate in response to short (few months) to medium
length (several years) events. In monthly-averaged data, the 26.0 isopycnal surface does not outcrop at
OSP, although the data record does show surface waters briefly reaching this density in the winters of
1959, 1961, 1965 and 1971. In the past 35 years, the 26.0 surface has not been observed at the ocean surface
at OSP, even with Argo profilers frequently sampling waters in this region since 2001. Waters of the Alaska
Gyre (centred at about 53N, 150W; see Fig. 3) ventilate to the 26.2 isopycnal during cool winters such as
2006 and 2007. The 26.5 and 26.7 surfaces have remained at fairly constant depths of 137 ± 17 and
168 ± 17 m. However rh = 26.9 and 27.0 show a deepening of 30 m over 50 years, with the result that a
greater volume of water now resides between the 26.7 and 26.9 surfaces.
It is not immediately apparent, when describing temporal trends in the ocean interior, whether it is appropriate to track constant depth, density or salinity surfaces. Each has its merits: depth indicates changes in habitat range for marine organisms, density is essential in describing transport within the ocean, and salinity
follows the most conservative property of oceanic waters. By choosing to follow isopycnal (density) surfaces,
changes in the interior ocean will deviate slightly from salinity. For example, a temperature increase at OSP on
the 26.7 isopycnal of 0.6 C over 50 years (Fig. 4) requires waters to become more saline by 0.01, causing a
density surface deepening of 1.5 m (a trivial amount) compared with salinity.
Rates of T and O2 change are computed for depth, density and salinity surfaces at OSP (Table 1; density
trends shown in Fig. 4). Trends listed were found significant at the 95% confidence level, with confidence levels
being estimated using a standard t-test but with the number of degrees of freedom reduced to allow for autocorrelation as outlined in von Storch and Zwiers (2002). Warming rates are as high as 0.018 C y1 at 200 m,
and oxygen decreases by as much as 0.96 lmol kg1 y1 at 150 m. Rates of change diminish below 200 m, but
remain detectable at 1000 m.
It is interesting that if we restrict ourselves to the occasional ship-board sampling at OSP then the warming
trend in the mixed layer reported by Freeland et al. (1997) is no longer significant at the 95% confidence level.
However, if we examine the so-called Reynolds OI.v2 data set for this region (Reynolds and Marsico, 1993)
then their monthly data from 1982 to present produces a warming trend in mid-winter (January through
March) SST observations of 0.020 C y1 (95% significance level of 0.008 C y1). The Reynolds data show
a rather weaker and less significant warming trend in mid-summer SSTs. This would suggest that a warming
trend is present even though it may not be convincingly apparent in direct observations at OSP. Crawford
et al. (2007) observe a very similar warming trend along the survey line from OSP to the British Columbia
coast of 0.9 C over 48 years, but note that it is only significant at the 90% confidence level. Because both
the upper and intermediate ocean are warming at similar rates, an increasing stratification of the upper ocean
is the result of a freshening mixed layer (salinity decrease of 0.0036 y1 between 1956 and 2006, similar to
0.0043 y1 previously reported for OSP by Freeland et al., 1997).
Oxygen trends in the mixed layer cannot be reliably assessed because strong seasonality, driven by both solar
radiation and biology, produces an annual cycle of 70–80 lmol kg1. Our data suggest oxygen is decreasing at
Latitude (N)
60
50
40
30
240
230
220
210
200
190
180
170
160
150
140
130
120
Longitude (W)
Fig. 3. Surface density (rh) in March 2006, · marking the locations of reporting Argo profilers. The region with densities higher than 26.5
is found along the western margin of the subarctic Pacific. Density contours are every 0.2, except 26.5 has been added. Note that there are
few or no floats in marginal seas of the North Pacific, therefore contours should be disregarded in these regions.
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
185
7
2
1
Temperature (ºC)
6
5
4
3
300
Oxygen (μmol kg-1)
250
200
150
100
50
2004
2000
1996
1992
1988
1984
1980
1976
1972
1968
1964
1960
1956
0
Fig. 4. Temperature and oxygen trends at Ocean Station P on the 26.5 (·), 26.7 (e), 26.9 (+) and 27.0 (h) isopycnal surfaces and at
station P4 (n) on the 26.7 surface. T and oxygen trends from linear regressions are provided in Table 1. Depth ranges (average and
standard deviation) are 140 ± 15 m, 168 ± 17 m, 278 ± 27 m and 370 ± 44 m on the 26.5, 26.7, 26.9 and 27.0 isopycnals, respectively. P4 is
warming at 0.0084 C y1, with O2 declining at 1.22 lmol kg1 y1. Two mesoscale eddies are labelled 1 and 2.
Table 1
Rates of change on various depth, density and salinity (33.25–34.00) surfaces at OSP based on linear regressions for data between 1956 and
2006
Surface
T (C y1)
O2 (lmol kg1 y1)
Surface
T (C y1)
O2 (lmol kg1 y1)
ML
150 m
200 m
300 m
400 m
600 m
800 m
1000 m
4000 m
ns
+0.016
+0.018
+0.013
+0.010
+0.0059
+0.0029
+0.0020
ns
ns
0.96
0.57
0.18
0.07
0.07
0.18
0.22
0.08
26.3rh
26.5rh
26.7rh
26.9rh
27.0rh
33.25
33.50
33.75
34.00
0.000
0.009
0.012
0.011
0.008
0.0083
0.0076
0.016
0.0094
0.54
0.70
0.68
0.60
0.39
0.57
0.64
0.07
0.10
Non-significant (ns) trends are found for surface temperature due to high seasonal variability and for 4000 m temperature due to the low
precision (0.01 C) of early data. Salinity trend for the mixed layer (ML) is 0.0036 y1.
186
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
0.17 lmol m2 y1, but this analysis is biased by a lack of late winter data (March–May) since 1982. A declining
trend might be anticipated due to the reduced solubility of oxygen in warming waters.
Below the mixed layer, seasonality has little effect on temperature or oxygen. At all depths below the mixed
layer to at least 1000 m, oxygen is declining and temperature increasing at OSP (Table 1). Oxygen is decreasing
more quickly at 800–1000 m than in waters in the 400–600 m range, suggesting more organic carbon remineralization in this layer. At 4000 m, a weak trend towards lower oxygen is evident whereas temperature data
are not precise enough to detect a trend (pre 1991 T data being reported only to 0.01 C). Whitney and Freeland (1999) previously reported that OSP waters were warming to at least 1000 m, with the strongest warming
occurring at 200 m.
Four isopycnal surfaces at OSP are selected for trend analyses (Fig. 4). The 26.5 and 26.7 surfaces show
very similar patterns over time, with temperatures rising from 4.4 C in the early record to 5.0 C at present.
Linear regressions through all data indicate a warming rate of 0.009 and 0.012 C y1 on these two density
surfaces, at ocean depths of between 120 and 200 m. The 26.9 and 27.0 isopycnals, found at depths between
230 and 420 m, show warming trends of 0.011 and 0.008 C y1. These trends are strongly influenced by the
cool (less saline) 1960s, so much so that warming is not apparent since 1972. Also plotted are 26.7rh data from
a Line P station (P4) on the continental slope, these results showing ocean warming by 0.008 C y1 and oxygen declines of 1.22 lmol kg1 y1. The California Undercurrent is a major component of these slope waters
(Mackas et al., 1987), so trends will be strongly influenced by variability to the south. Temperature and oxygen
at P4 are similar to values observed for mesoscale eddies at OSP in 1960 and 1974.
Oxygen levels at OSP decline over time by rates ranging from 0.39 to 0.70 lmol kg1 y1 (Fig. 4). Lower
oxygen levels are evident during the cool periods of the 1960s and 1999–2002, and brief periods of oxygen
increase are occasionally associated with abrupt warming, (late 1950s, mid 1970s and 2004). In the eastern
subarctic Pacific, oxygen levels increase within warm, saline waters from the subtropics, decrease in cool, fresh
waters from the Alaska Gyre, and strongly decrease in warm, fresh waters carried by mesoscale eddies.
For the period 1987–2005, OSP oxygen (O2) and nitrate (NO3) were integrated between 100 and 600 m
(Fig. 5). As expected, their trends are mirror images of each other. Between 1994 and 2003, oxygen declined
at a rate of 2.4 mol m2 y1 as nitrate increased at 0.26 mol m2 y1, with the oxygen decline to nitrate
increase ratio being 9.2. Between 2003 and 2005, the trend reversed with oxygen increasing (6.7 mol m2 y1)
and nitrate declining (0.66 mol m2 y1) yielding an O2/NO3 ratio of 10.2. Average depth integrated values
(±1 SD) are 19.2 ± 0.7 mol NO3 m2 and 45.8 ± 6.5 mol O2 m2, with oxygen concentrations averaging from
290 lmol kg1 at 100 m to 30 lmol kg1 at 600 m and nitrate varying between 17 lmol kg1 at 100 m and
44 lmol kg1 at 600 m.
22
60
55
50
20
45
40
19
Nitrate (mol m-2)
Oxygen (mol m-2)
21
35
18
30
17
2006
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
25
Fig. 5. Nitrate (n) and oxygen (e) integrated between 100 and 600 m at Ocean Station P. Rates of change from 1994 to 2003 are
+0.26 mol NO3 m2 y1 and 2.4 mol O2 m2 y1. Between 2003 and 2006, rates are 0.66 mol NO3 m2 y1 and +6.7 mol O2 m2 y1.
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
187
3.2. Oxygen distribution on isopycnal surfaces from WOCE one time surveys
To interpret trends at OSP, it is necessary to identify the regions in the North Pacific where the interior
waters of the subarctic region exchange gases with the atmosphere. WOCE completed its surveys of the subarctic Pacific (Fig. 6) between 1985 (P1) and 1994 (P15). These data are used to characterize waters on the 26.5,
26.7 and 26.9 isopycnal surfaces poleward of 35N, the lightest of these layers commonly outcropping in the
western Pacific during winter (Mecking et al., 2006, Fig. 3). As well as showing WOCE stations, Fig. 6 also
identifies sites at which oxygen saturation exceeded 95% on the 26.5 isopycnal, these locations being areas
where winter ventilation of interior subarctic waters is inferred. Three larger coastal regions are also encircled,
identifying areas that will subsequently be shown to have low.
Each of the selected density surfaces deepen from west to east in the subarctic (Fig. 7), the densest layer being
200 m in the vicinity of 160E and 460 m near North America. As subarctic waters flow across the Pacific in
the NPC, they warm and lose oxygen. Coldest waters on the 26.5 isopycnal (1.5 to 1.5 C, data from Okhotsk
not plotted) are found in the Sea of Okhotsk and along the coasts of northern Japan and Russia. South of the
Alaska Gyre NPC waters are between 6.8 and 8 C. Waters >8.0 are found south of 42N in the subtropical
domain. Near the North American coast, waters have warmed by 2–6 C and lost 35–200 lmol kg1 oxygen
in their 6000 km journey across the Pacific. The lowest oxygen levels are found in coastal waters of the
California Undercurrent, decreasing to 45 lmol kg1 on the 26.9 isopycnal at a depth of 460 m.
On the 26.5 isopycnal surface oxygen levels >95% saturation are broadly found in Western Subarctic Gyre
(Figs. 6 and 8), also on the shelf of the Okhotsk Sea, in the Bering Sea Gyre and near the Russian coast in the
East Kamchatka Current. Many of the WSG values come from the 1985 survey of P1, not from the warm
1990s when other lines were surveyed. Each of the WOCE surveys was carried out in summer when thermal
stratification prevented the 26.5 isopycnal from reaching the ocean surface.
Oxygen decreases to saturation levels of 65% in the eastern waters of the NPC. On the 26.7 (26.9) isopycnals, oxygen declines from >80% (50%) saturation in the Okhotsk Sea and near Japan to <30% (10%) along
the North American coast from California to British Columbia. Low oxygen and high nitrate occur in coastal
regions along: (1) the Aleutian Islands on section P14, (2) the south coast of Alaska and (3) the California to
British Columbia coast (Fig. 8). Region 2 includes the waters of the Alaska Gyre. Between region 2 and the
North Pacific Current directly to the south on the 26.5 isopycnal, oxygen increases from 150 to
250 lmol kg1 as nitrate decreases from 32 to 20 lmol kg1 (O2/NO3 ratio of 9.2). OSP lies in the midst
of this gradient.
Normally, nitrate distribution will be the inverse of oxygen, since oxygen is consumed in a fairly constant
ratio as nitrate is remineralized from organic N in seawater. In general, this is evident in Fig. 8. High oxygen
60
Latitude (N)
56
2
P1W
52
1
48
44
40
36
140
P15
P13
P1
3
P16
P14
P17
160
180
200
220
Longitude (E)
Fig. 6. Station locations of WOCE one-time surveys in the subarctic Pacific (lines P1 through P17). Regions of low oxygen (ovals
numbered 1–3) are referred to in subsequent figures. Stations highlighted by grey circles in the western subarctic region indicate locations
at which oxygen saturation exceeds 95% on the 26.5 isopycnal.
188
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
Longitude (E)
150
170
190
210
230
250
170
190
210
230
250
0
Depth (m)
100
200
300
400
500
8
Temperature (ºC)
7
6
5
4
3
2
1
0
Oxygen (μmol kg-1)
400
300
200
100
0
150
Longitude (E)
Fig. 7. Plots of: (a) depth, (b) temperature and (c) oxygen on the 26.5 (+), 26.7 (d) and 26.9 (e) isopycnal surfaces between 45 and 48N,
155E and 125W from WOCE surveys. Data in panels a and c are fit with a third order polynomial to show trends (dashed lines).
and low nitrate is found in waters near the Asian coast. However, nitrate is not highest off the coast of California, despite oxygen levels being lowest for the entire subarctic Pacific. Except for one high nitrate value off
the BC coast, nitrate reaches maximum concentrations of 33 lmol kg1 along the south Alaska coast. The
high BC value (37 lmol kg1) is the result of estuarine outflow from a coastal channel (Whitney et al.,
2005). Between P13 and P14, nitrate drops from 18 to 14 lmol kg1 in the latitudinal band 42–45N, its lowest concentration in the subarctic Pacific, excluding the Sea of Okhotsk.
The covariance of nitrate and oxygen is seen more clearly in Fig. 9. Within the scatter of data, regional
patterns emerge. Data are bound by dashed lines with O2/NO3 slopes of 9.2, the same ratio observed along
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
189
400
100
Oxygen (μmol kg-1)
300
80
60
200
1
100
P1W
P1
P13
P14
P15
P16
2
3
P17
0
-2
0
2
4
6
8
10
Temperature (ºC)
40
2
1
Nitrate (μmol kg-1)
3
30
20
10
0
-2
0
2
4
6
8
10
Temperature (ºC)
Fig. 8. Oxygen vs. temperature (top) and nitrate vs. temperature (bottom) for all data on the 26.5 isopycnal surface from WOCE survey
lines. Oxygen (salinity 33) is shown for 60%, 80% and 100% saturation (dashed lines). Clusters of oxygen and nitrate data are circled (low
O2) and their locations shown in Fig. 6.
the 26.5 isopycnal in the eastern subarctic. Data along the California and southern BC coast cluster along the
lower line (oval 3). Successive data clusters off the south coast of Alaska and through the Aleutian Islands
(ovals 1 and 2) show higher oxygen and nitrate as waters take on more subarctic characteristics. Data clusters
with the lowest nitrate levels outside the Sea of Okhotsk are found in subtropical waters on WOCE sections
P13 and P14.
4. Discussion
4.1. Oxygen and temperature in the interior ocean at OSP
A freshening of the surface layer and shoaling of the winter mixed layer from 120 to 100 m between the
1960s and 1990s (Freeland et al., 1997) are symptoms of continually increasing stratification of the upper
ocean at OSP. Royer and Grosch (2006) also find that upper ocean stratification is enhancing ocean stratification along the Alaska coast. As a result, nutrient transport to the mixed layer weakened along Line P
between the 1970s and 1990s (Whitney et al., 1998). The 50 years of observations at OSP suggests upper ocean
stratification is also reducing oxygen transport to the ocean interior. The most rapid rates of O2 decrease are
190
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
40
1
2
35
3
Nitrate (μmol kg-1)
30
25
20
15
10
5
P1W
P1
P13
P14
P15
P16
P17
0
0
100
200
300
400
Oxygen (μmol kg-1)
Fig. 9. Nitrate vs. oxygen for WOCE data on the 26.5 isopycnal. Dashed lines show an O2:NO3 ratio of 9.2 for waters of subtropical
(lower) and subarctic (upper) nature. Three circled data groups are the same as those in Figs. 6 and 8, being found: (1) near the Aleutian
Islands, (2) along the Alaska south coast and (3) along the BC to California coast.
observed between 100 and 400 m, on density surfaces of 26.3–27.0. Over this range, oxygen is decreasing on
average by 123 mmol m2 y1, equivalent to a loss of 22% over 50 years. This loss of oxygen has resulted in
the ‘‘hypoxic’’ boundary (considered to be 60 lmol kg1; Gray et al., 2002) shoaling from 400 to 300 m
between the 1950s and the present.
Oxygen declines are detectable to at least 1000 m at OSP. The vertical transport of particulate organic carbon at this site is 6.6 g C m2 y1 to 200 m and 2.7 g C m2 y1 to 1000 m (Wong et al., 1999). Carbon loss
between these depths (2.9 g C = 240 mmol C) is greater than the 50 year average integrated rate of oxygen
loss (165 mmol O2 m2 y1 from Table 1). Reduced ocean ventilation results in carbon dioxide being stored
in the ocean interior at a rate we calculate to be 130 mmol m2 y1 in the vicinity of OSP (O2 consumption/1.3
to convert to CO2; Anderson, 1995). This accumulation rate for CO2 is considerably less than the average carbon remineralization rate of 2000 mmol m2 y1 estimated for subsurface waters (200–900 m) of the subarctic Pacific by Feely et al. (2004) or for the integrated oxygen consumption rate (150–800 m) derived by Imasoto
et al. (2000) of 1600 mmol m2 y1 for the same region, and suggests perhaps 5–10% of the CO2 being produced by remineralization of organic detritus is being stored in the ocean interior. Without oxygen transport
to OSP, the approximately 34 mol O2 m2 found between 200 and 1000 m (Table 1) would be consumed
within 15–20 years.
Warming is evident in subsurface waters over the 1956–2006 period, but without the strong cool period in
the 1960s, this trend would weaken considerably. During the cool 1960s and from 1995 to 2003, oxygen levels
diminish relatively rapidly. Andreev and Baturina (2006) suggested this was the result of an intensified winter
Aleutian Low pressure system, causing a strengthening and expansion of the Alaska Gyre. However, Crawford et al. (2007) show that increased southward transport of cool waters in this region also occur when the
positioning of the Aleutian Low and North Pacific High create stronger westerly winds towards the BC coast.
The warm period in the early 1990s, associated with a series of El Niño events, slowed the rate of oxygen
decline on the 26.5 and 26.7 isopycnals but had little effect on deeper waters.
Only in the past decade has it been clear that mesoscale eddies transport coastal waters westward against
prevailing currents in the NE Pacific (Whitney and Robert, 2002). Along the BC and SE Alaska coasts, waters
are warmer and contain less oxygen than in the Alaska Gyre, due both to the influence of the California
Undercurrent and to the increased oxygen demand fuelled by high primary productivity. During the Weathership era, water properties were measured frequently enough to resolve the passage of eddies through OSP.
Both in 1960 and 1974, depressed isopycnals, elevated temperature and low oxygen signal the presence of these
anticyclonic eddies at OSP for 4–6 months. In these eddies, temperature and oxygen levels on the 26.7 isopycnal are similar to those measured at station P4 near the BC coast (Fig. 4), indicating waters of coastal origin
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
191
that contain a substantial CUC component. If these short duration anomalies were due to enhanced recirculation of waters from the coast of Alaska to OSP (as described in Whitney et al., 2005), depressed isopycnal
surfaces coincident with the low oxygen/high temperature would not accompany the anomaly, nor would the
temperature anomaly be as strong (Fig. 8). Some of the variability seen over the past couple of decades could
also have resulted from such eddies.
Oxygen has been observed oscillating in the western waters of the subarctic Pacific on an 18–20 year cycle
(Ono et al., 2001; Andreev and Kusakabe, 2001; Watanabe et al., 2003). A recent suggestion is that this oscillation is driven by the 18.6 year lunar nodal cycle (Yasuda et al., 2006; Andreev and Baturina, 2006) which
causes diurnal tidal amplitude to vary by 20% and consequently gives rise to strong tidal mixing variability
in the narrow passes between Okhotsk Sea and the Pacific Ocean. On the 26.9 and 27.0 isopycnals at OSP,
such an oscillation can be detected, with oxygen maxima appearing in 1959, 1978 and 1995. These peaks occur
6–7 years after those observed off Japan and 9–10 years after the nodal maxima (1951, 1969 and 1988). Such
oscillations make the interpretation of climate trends difficult with short term data records.
4.2. Oxygen transport within the subarctic Pacific
Oxygen enters the interior ocean in areas where dense waters exchange gases with the atmosphere. In the
subarctic Pacific, the densest isopycnal surfaces (26.6) are mixed to the surface in the cold waters along the
Asian coast and in the Bering Sea. In the eastern Pacific, the 26.2 isopycnal has outcropped occasionally in the
Alaska Gyre in the past several years but at OSP, the 26.0rh has not been observed at the ocean surface since
1971 despite the frequent measurements that have being taken in this region by profiling Argo floats since 2001
(Freeland and Cummins, 2005).
Rapid declines in oxygen (60 lmol kg1 over 10 years; Watanabe et al., 2001; Kumamoto et al., 2004)
have been reported for the interior waters of the subarctic Pacific in recent years from repeat survey sections.
These surveys sampled through a period of persistently weak ventilation which resulted in high apparent oxygen utilization (Ono et al., 2001; Mecking et al., 2006; Andreev and Baturina, 2006, Fig. 4). Such periods provide a good opportunity to estimate oxygen consumption rates, since periods between sampling can be similar
to the aging of these water masses. Estimates of oxygen utilization rates vary from 2.1 M m2 y1 between
26.5 and 27.2rh in the NPC (Aydin et al., 2004) to 4.1 M m2 y1 over the depth range 200–900 m as an average for the subarctic Pacific (Feely et al., 2004). In comparison, the observed rate of oxygen decline at OSP was
2.4 M m2 y1 (100–600 m, Fig. 5) through the low ventilation period of 1994–2003, but is only
0.14 M m2 y1 over the same depth range (Table 1) when averaged over 50 years. Over the entire observation
period, annual oxygen consumption only slightly exceeded transport, but resulted in a 22% decline in oxygen
levels.
Circulation models developed in recent years clarify transport processes in the interior subarctic Pacific,
especially within the North Pacific Current. The strong Kuroshio and Oyashio Currents initiate the NPC near
the Asian coast. Ueno and Yasuda (2003) picked the 26.7 and 27.2 isopycnals to show how flow and mixing
along these surfaces transports oxygen from the ventilation regions along the Asian coast and heat from the
subtropics towards North America. Their model provided estimates of transport times from the edge of ventilated waters to OSP of 7 years on the 26.7 surface and 15 years on 27.2rh. A 10 year lag was observed in a
high oxygen, low temperature signal crossing from the WSG to the AG in the 1960s (Andreev and Watanabe,
2002). These results are consistent with the increase in age of NPC waters observed over a shorter section of
the NPC (between WOCE lines P14 and P17, Fig. 6) of from 4 years on 27.0rh to 6 and 8 years on the 27.1 and
27.2 isopycnals (Aydin et al., 2004).
Oxygen declines can be due to either increased biological demand or changes in physical processes. In modelling oxygen transport within the North Pacific, Deutsch et al. (2006) found little evidence of increased biological activity, but instead showed that changes in ocean circulation and ventilation accounted for both
episodic oxygen increases in the subtropics and declining oxygen in the subarctic Pacific. Reduced ventilation
of the 26.6 isopycnal was thought to be the major reason for decreasing oxygen levels in the subarctic region.
A freshening upper ocean (Freeland et al., 1997; Wong et al., 2001; Royer and Grosch, 2006), something we
continue to see at OSP, may be exacerbating this trend. Over 50 years, Andreev and Watanabe (2002)
observed a surface salinity decrease in the Bering Sea of 0.32 per century which is close to the 0.36 per century
192
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
we calculate for OSP. Surface freshwater transport from the eastern to western Pacific, which is strengthened
during periods of intense Aleutian Lows, appears to be an essential modulator of ocean ventilation throughout the subarctic region (Andreev and Baturina, 2006).
4.3. Using NO to characterize water masses influencing the subarctic Pacific
Broecker (1974) derived a conservative tracer using nitrate and oxygen to track water masses in the Atlantic, using the formulation 9NO3 + O2 = NO. Based on rates of change in oxygen and nitrate at OSP through a
period of persistent oxygen decline (Fig. 5) we slightly modify the NO parameter to be 9.2NO3 + O2. NO is a
water mass signature that is carried from its formation region; waters exchanging gases with the atmosphere in
cold regions with high nitrate levels will have high NO whereas warm waters of low nitrate produce a low NO
value. As waters sink away from their formation regions, remineralization of organic matter consumes oxygen
and produces nitrate in a fairly constant ratio as long as oxygen is available. Biotic processes in oxygen rich
waters do not alter the NO of a water mass, hence this parameter is considered conservative (does not alter
with time). When oxygen nears zero, denitrification by bacteria will begin to remove nitrate, producing anomalously low NO.
Based on Redfield ratios of major elements in marine algae, the O2/NO3 ratio is 8.5 during oxidation of
organic matter (e.g. Anderson, 1995). However, the carbon to nitrogen ratio of marine detritus at OSP
increases from 6.5 in the suspended particles of the photic zone to 8.0 in particles sinking to 100 m (Wong
et al., 1999), potentially increasing the O2/NO3 ratio to 9.5 for complete oxidation of this organic material.
For subsurface waters, a ratio of 9.2 fits well with the composition of its detrital material.
In the North Pacific, NO provides insights to water mass transport and mixing that temperature and salinity cannot. Strong regional differences are seen throughout the North Pacific (Fig. 10), with values up to
580 lmol kg1 being found on the 26.5 isopycnal in the Bering Sea during the WOCE survey in 1992 and
1993. Unpublished data from surveys we carried out along 165W in 1991 and 1992 also show NO values
approaching 580 lmol kg1 in the Western Subarctic Gyre (Whitney, unpublished). Minima for NO occur
across the subtropics and in the California Undercurrent. Okhotsk Sea Intermediate Waters differ from those
NO (μmol kg-1)
250
25.0
350
450
NO (μmol kg-1)
550
250
350
450
550
OSP
BSG
WSG
25.5
ST
AG
ST
Sigma theta
26.0
OS
26.5
27.0
CUC
27.5
W
E
28.0
Fig. 10. NO vs. density for representative profiles in the western (W) and eastern (E) Pacific. Western water types include subtropical (ST),
Okhotsk Sea (OS) where waters flow into the Pacific, Bering Sea (BSG) and Western Subarctic Gyres (WSG). Eastern water types are the
California Undercurrent (CUC), subtropics (ST), Ocean Station P (OSP) and Alaska Gyre (AG). Dashed horizontal lines highlight
isopycnal surfaces of interest.
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
193
in the EKC and WSG either because the 26.5 and 26.7 isopycnals are close enough to the surface on the northern shelf that nitrate can be utilized by phytoplankton or due to denitrification within sediments of the broad
shelf (Chen et al., 2004; Yoshikawa et al., 2006). Perhaps other processes such as brine rejection during ice
formation, and incomplete ventilation of waters under winter ice contribute to the low NO signature of OSIW.
In the eastern subarctic region, Alaska Gyre waters contain the highest NO values, although these waters only
ventilate to the 26.2 isopycnal. No subsurface maximum indicative of winter ventilation to depth is seen in
the profiles selected from 1993 WOCE surveys. Situated between the NPC and the gyre, OSP has an intermediate nitrate–oxygen value compared with AG and the subtropics. Low NO in the California Undercurrent is
the result of denitrification further south (Castro et al., 2001) and can be used to track its influence up the
coast of BC and Alaska. Especially the denitrification signatures of Okhotsk and the CUC are lost when trying
to use temperature or salinity to track oxygen transport and consumption in the interior ocean.
Other NO values from the North Pacific covary with temperature (Fig. 11). Trends are similar on three isopycnals, except there is less evidence of the well ventilated BSG/EKC water below the 26.5 isopycnal. Cold
OSIW is still common on the 26.7rh, however Okhotsk Sea waters are warmer (1.4 C) and have lower
NO (465 lmol kg1) on the 26.9 surface. Such low values at depth suggest a strong shelf component to these
waters (Wong et al., 1998) and less contribution from the EKC. Assuming 100% oxygen levels during water
formation, preformed nitrate concentrations vary between 7 and 27 lmol kg1 north of 35N in the Pacific.
Data from OSP over the period 1987–2006 all fall in the centre of the mixing line for 26.7rh waters, whereas
NO from the near coastal station P4 shows similarity to the CUC. The distinct characteristics of various contributing water masses to the NPC make them useful in defining their influence on this trans-Pacific current
and within the Gulf of Alaska.
Water mass characteristics are summarized in Table 2 and mapped out for the 26.7 isopycnal in Fig. 12.
There is a clear mixing line (Fig. 11) between the cool, NO rich waters of the western subarctic (2 C and
BSG/WS
600
550
25
NO (μmol kg-1)
OSIW
500
OSP
450
P4
ST
400
5
350
CUC
300
-2
0
2
4
6
8
10
Temperature (ºC)
Fig. 11. NO vs. temperature on the 26.5 (·), 26.7 (h) and 26.9 (m) isopycnal surfaces. Dashed lines represent preformed NO at 100%
oxygen saturation and nitrate concentrations of 5 and 25 lmol kg1 at a salinity of 33. Waters of the Bering Sea Gyre (BSG) and East
Kamchatka Current (EKC) show elevated NO values, whereas lowest values identify subtropical (ST) waters. Okhotsk Sea Intermediate
water (OSIW) has a slightly reduced NO signature. The CUC waters along the California coast show weak evidence of denitrification on
all density surfaces. Also shown are NO values for OSP (e) and P4 (s) on the 26.7 isopycnal for the period 1987–2006.
194
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
Table 2
Water mass characteristics (±SD of data in defined regions) on the 26.5 and 26.7 isopycnal surfaces from WOCE data
Water type
26.5 isopycnal
Bering Gyre, 53–56N
WSG at 165E, 47–52N
Okhotsk, 49–52N
NPC at 179E, 45–48N
NPC at 165E, 41–46N
NPC/AG at 150W 48–54N
Alaska coast, 135–140W 56–57N
Subtropics at 165E 35–41N
CUC at 124W, 38N
26.7 isopycnal
T (C)
Salinity
NO (lmol kg1)
T (C)
Salinity
NO (lmol kg1)
2.6 ± 0.1
2.1 ± 0.5
0.0 ± 0.5
3.0 ± 0.3
4.1 ± 0.2
4.3 ± 0.2
6.0 ± 0.2
8.2 ± 0.2
7.83 ± 0.08
33.22 ± 0.02
33.2 ± 0.1
33.03 ± 0.02
33.27 ± 0.10
33.4 ± 0.1
33.4 ± 0.1
33.65 ± 0.05
34.08 ± 0.05
34.00 ± 0.05
575 ± 5
545 ± 5
504 ± 12
505 ± 15
460 ± 10
450 ± 5
415 ± 5
380 ± 10
371 ± 2
3.64 ± 0.08
2.9 ± 0.7
0.35 ± 0.45
3.8 ± 0.7
4.4 ± 0.7
4.4 ± 0.4
6.0 ± 0.2
6.0 ± 0.5
6.9 ± 0.3
33.61 ± 0.03
33.52 ± 0.09
33.27 ± 0.06
33.64 ± 0.07
33.69 ± 0.11
33.70 ± 0.06
33.91 ± 0.02
33.93 ± 0.07
34.08 ± 0.04
482 ± 7
490 ± 20
508 ± 11
470 ± 16
450 ± 15
451 ± 10
417 ± 5
412 ± 10
380 ± 8
66
Latitude (N)
62
58
54
50
417
454+/-15
460
482
508
490
453+/-8
451
470
408+/-8
46
450
42
412
38
140E
160E
380
160W
180
140W
120W
Longitude
Fig. 12. NO characteristics (lmol kg1) of water masses (identified in Table 2) on the 26.7 isopycnal surface in the Northern Pacific Ocean.
Arrows show the direction of flow of major currents and gyres. Standard deviations are provided for three time-series stations in the NE
Pacific, all other values are based on one-time WOCE surveys.
NO 500 lmol kg1) and those of the subtropics (6 C, NO = 412 lmol kg1), although how the cool end
member forms is not obvious. Its contributors will be OSIW (NO 508 lmol kg1) and the WSG
(NO 490 lmol kg1). To estimate the influence of the western subarctic on the region around Ocean Station
P, we use an NO of 500 lmol kg1 to type this source region with an uncertainty of 10 lmol kg1. The
subtropics component is well defined along 165E. These waters will be somewhat contaminated by the subarctic as North Pacific Intermediate Water forms along the NPC (You, 2005). Still, they are the waters that
supply the subtropical component of the NPC.
We now use NO to estimate contributions of western subtropical and subarctic waters to the makeup of
water masses in the eastern subarctic. Table 3 summarizes these estimates and shows that, for the WOCE data
set of the 1990s, waters of the Alaska Gyre were 60% from the WSG region and 40% from the subtropics on
the 26.7 isopycnal. On the edge of AG, stations Z9 and OSP contained slightly more water from the subtropics
(Fig. 12), although here variability could be assessed. With fewer samplings in a period of high variability due
to a strong El Niño (1997/98) and La Niña (1999), the subtropical component at Z9 varied between 25% and
65%. At OSP, the variability (assessed as 1SD) was less, with the subtropics contributing 37–57%. An ‘‘outlier’’ for OSP in February 1995 (Fig. 13) suggested waters were 90% subtropical, with their abnormally warm
5.7 C temperature supporting this estimate.
An important issue along the coast of Oregon, Washington and BC in the past several years has been
hypoxia of shelf waters (Grantham et al., 2004). Understanding which waters (subarctic or subtropical) are
being upwelled onto the shelf in summer is core to determining why these events are now occurring when they
had not before. NO is a good parameter to track influences of the California Undercurrent or AG waters in
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
195
Table 3
Estimated contributors of oxygen along the 26.7 isopycnal surface in the NE Pacific
Water mass
% Subarctic
% Subtropical
Comments
WSG
ST
AG
Z9
OSP
100
0
60
55 ± 20
53 ± 10
0
100
40
45 ± 20
47 ± 10
1992
1992
1993
1991–1999, n = 8
1987–2006, n = 40
% AG
% CUC
Comments
P4
SE Alaska
35 ± 10
54
65 ± 10
46
1987–2006, n = 44
1993
480
470
NO (μmol kg-1)
460
450
440
430
420
410
1986
1990
1994
1998
2002
2006
6.0
Temperature (ºC)
5.5
5.0
4.5
4.0
3.5
1986
1990
1994
1998
2002
2006
Fig. 13. Changes in NO and temperature at OSP on the 26.5 (n), 26.7 (·) and 26.9 (h) isopycnal surfaces between 1987 and 2006.
this region. WOCE surveys suggest waters of the SE Alaska coast are a near equal mixture of CUC and AG
waters in 1993. At station P4 on the southern BC coast, CUC has comprised 55–75% of the slope waters on
the 26.7 isopycnal in the past 20 years. This is similar to the findings of Mackas et al. (1987) for summer 1980
when CUC formed a large fraction of the waters below 100 m along the continental slope of southern BC.
The interannual variability of NO and temperature at OSP are plotted for the period 1987–2006 (Fig 13).
Both parameters clearly show the influence of the serial El Niños in the mid 1990s. Temperatures on the 26.5
and 26.7 isopycnals rose to almost 6 C in 1995 as NO decreased as low as 420 lmol kg1. Assuming contributing water mass properties changed little through this period, this low NO suggests waters at OSP were 90%
196
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
subtropical. Such a strong change in water properties requires a substantial northward movement of the NPC.
With the advent of the Argo array of profiling floats, it has become possible to monitor shifts in large-scale
ocean current systems. Freeland and Cummins (2005) and Freeland (2007) demonstrated methods that allow
such current tracking to be executed and show that since the initiation of the Argo array in the NE Pacific in
2001 some large shifts (several degrees latitude) in the location of the NPC have indeed taken place. Because
Argo has been deployed for only a few years, the statistics available from such mappings are insufficient to
allow any shifts to be correlated with the Line P time series. However, the idea that changes in water properties
might be associated with spatial variability in major ocean currents is supportable as shifts in the NPC are now
known to take place.
4.4. Implications of oxygen decline
Understanding ocean ventilation provides great insight into the circulation of the interior ocean. Because
ventilation sites are rare, processes within them have the potential to control biological activity over broad
areas and time scales. Past periods of weak ventilation appear to be the cause of near anoxic conditions in
regions affected by the subarctic Pacific. Core records from Baja California show periods of near anoxia thousands of years ago (Van Geen et al., 2006), the suggested cause being weakened ocean ventilation in the western subarctic Pacific. WOCE repeat sections consistently point to a rapid decline in oxygen throughout the
subarctic Pacific between the late 1980s and 2000, reporting oxygen declines of 40–80 lmol kg1 in large
patches of the subarctic over periods of 6–14 years (Watanabe et al., 2001; Emerson et al., 2004; Kumamoto
et al., 2004; Mecking et al., 2006). Such rates of decline, if persisting, could turn broad regions of the subarctic
Pacific hypoxic within a decade. However, observations at OSP show there are multi-year periods of virtually
no ventilation such as those seen in the late 1960 and 1990s (Fig. 4), followed by periods of oxygen enrichment
that may be associated with increased tidal mixing amongst the Kuril and Aleutian Islands (Yasuda et al.,
2006; Andreev and Baturina, 2006). Enhanced mixing at nodal tidal maxima in 1969 and 1988 match periods
of minimum AOU (maximum oxygen concentration) in Oyashio waters off northern Japan (Ono et al., 2001).
With a lag of 6–9 years, this signal is evident on the 26.9 isopycnal at OSP and perhaps on shallower surfaces
as well (Fig. 4). The broad outcropping of the 26.5 and 26.6 isopycnals in March 2006 (Fig. 3) may be the
result of increased tidal mixing as the next nodal maximum is approached. Andreev and Kusakabe (2001)
attempted to link oscillations in oxygen levels in the Western Subarctic Gyre and Sea of Okhotsk with atmospheric forcing. They observed oxygen maxima on the 27.0 isopycnal surface in WSG near 1951, 1969 and
1988, each a nodal tidal maximum.
In the depth range 125–300 m at OSP, oxygen decreased between 17% and 30% (20–40 lmol kg1) from
1956 to 2006. The oscillations recorded at OSP have resulted in the hypoxic boundary (60 lmol kg1) varying between 400 and 250 m depth. Such changes must affect the distribution of organisms intolerant of low
oxygen. Little research appears to have been done on the distribution of oceanic species within oxygen gradients, however quite a sharp oxygen boundary of 3.4 mg l1 (100 lmol kg1 or 28% O2 saturation) was noted
for Atlantic cod in the Gulf of St Lawrence (D’Amours, 1993), demonstrating oxygen is a crucial parameter in
defining ocean habitat. Marine animal life depends on oxygen. The impacts of oxygen declines need to be
understood for both open ocean and coastal ecosystems. Low oxygen is of broad concern because of human
impacts on continental margins where increased waste discharges place demands on the oxygen supply. This
can arise from increased nutrient inputs such has been observed in the Gulf of Mexico (Turner and Rabalais,
1994) or localized discharges from coastal industry and municipal sewage treatment plants that may affect
areas of restricted circulation. More ominously, hypoxia may increase in the subarctic Pacific due to global
warming as upper ocean stratification strengthens (Freeland et al., 1997; Royer and Grosch, 2006). Bakun
(1990) theorized that a warming world would establish stronger thermal gradients between land and ocean,
resulting in stronger along shore winds. This would enhance upwelling, increasing nutrient supply to surface
waters and placing higher demands on bottom water oxygen as an increased flux of detritus is consumed. Not
imagined by Bakun was the decline in oxygen and increase in nutrient levels in the ocean interior. Increased
upwelling in his scenario will be accompanied by transport of waters predisposed to cause hypoxia or anoxia.
Fish and crab kills in the Oregon upwelling region in the past few years (Grantham et al., 2004) suggest this is
occurring to some extent already. As oxygen declines in the subarctic Pacific, the hypoxic threshold will shoal.
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
197
For example on the 26.7 isopycnal at OSP, oxygen levels have declined from 187 to 150 lmol kg1 over 50
years of observations. With no ventilation and oxygen consumption rates of 4 lmol kg1 y1 (Feely et al.,
2004), it would take a little more than 20 years to create hypoxia (60 lmol kg1) in these waters. This isopycnal is less than 250 m deep near the North American coast, a region where increased oxygen demand from
productive waters places a high demand on oxygen.
On the Oregon to British Columbia coast, upwelling draws waters from depths of 100 to >250 m (Freeland
and Denman, 1982; Whitney et al., 2005). If coastal upwelling does strengthen, and with some assurance that
oxygen levels will continue to decline (and nutrient levels increase) in the subarctic waters of the Pacific, it is
reasonable to project that shelf and slope ecosystems will lose oxygenated habitat. The few fish species such as
sablefish and some rock fishes that tolerate low oxygen may expand their territory, but in general mid water
organisms will be forced to find shallower habitat or perish. This will increase competition for resources and
may expose some species to greater predation from above. In coastal basins and fjords whose basin waters are
rejuvenated with periodic replacement from the ocean, declining oxygen levels in the subarctic Pacific could
lead to serious hypoxia resulting in widespread mortality of benthic species.
Acknowledgements
Collecting 50 years of data without introducing biases and variability that obscures climate signals takes the
dedication of many field and lab personnel; most have been our friends and sea-going comrades for decades.
Noteworthy in compiling data from OSP are Sus Tabata (T/S), Bernard Minkley (oxygen), Janet BarwellClarke (nutrients) and Joe Linguanti (data archiving). Encouraging comments and suggestions from reviewers
and from Denis Gilbert are gratefully acknowledged. This paper was presented at the PICES/DFO Symposium on ‘‘Time Series of the N.E. Pacific: A symposium to mark the 50th anniversary of Line P’’ in Victoria,
BC, Canada, July 2006.
References
Anderson, L.A., 1995. On the hydrogen and oxygen of marine phytoplankton. Deep-Sea Research I 42, 1675–1680.
Andreev, A.G., Baturina, V.I., 2006. Impacts of tides and atmospheric forcing variability on dissolved oxygen in the subarctic North
Pacific. Journal of Geophysical Research 111. doi:10.1029/2005JC003.
Andreev, A.G., Kusakabe, M., 2001. Interdecadal variability in dissolved oxygen in the intermediate water layer of the Western Subarctic
Gyre and Kuril Basin (Okhotsk Sea). Geophysical Research Letters 28, 2453–2456.
Andreev, A., Watanabe, S., 2002. Temporal changes in dissolved oxygen of the intermediate water in the subarctic North Pacific.
Geophysical Research Letters 29, 1680. doi:10.1029/2002GL015021.
Aydin, M., Top, Z., Olson, D.B., 2004. Exchange processes and watermass modification along the subarctic front in the North Pacific:
oxygen consumption rates and net carbon flux. Journal of Marine Research 62, 153–167.
Bakun, A., 1990. Global climate change and intensification of coastal ocean upwelling. Science 247, 198–201.
Barwell-Clarke, J.E., Whitney, F.A., 1996. Institute of Ocean Sciences nutrient methods and analysis. Canadian Technical Report of
Hydrography and Ocean Science 182, 43.
Broecker, W., 1974. ‘‘NO’’ a conservative water-mass tracer. Earth and Planetary Science Letters 23, 100–107.
Carpenter, J.H., 1965. The Chesapeake Bay Institute technique for the Winkler dissolved oxygen method. Limnology and Oceanography
10, 141–143.
Castro, C.G., Chavez, F.P., Collins, C.A., 2001. Role of the California Undercurrent in the export of denitrified waters from the eastern
tropical North Pacific. Global Biogeochemical Cycles 15, 819–830.
Chen, C.-T.A., Andreev, A.G., Kim, K.-G., Yamamoto, M., 2004. Roles of continental shelves and marginal seas in the biogeochemical
cycles of the North Pacific Ocean. Journal of Oceanography 60, 17–44.
Crawford, W., Galbraith, J., Bolingbroke, N., 2007. Line P ocean temperature and salinity, 1956–2005. Progress in Oceanography 75,
161–178.
Curry, R., Dickson, B., Yashayaev, I., 2003. A change in the freshwater balance of the Atlantic Ocean over the past four decades. Nature
426, 826–829.
D’Amours, D., 1993. The distribution of cod (Gadus morhua) in relation to temperature and oxygen level in the Gulf of St. Lawrence.
Fisheries Oceanography 2, 24–29.
Davis, J., 1975. Minimal dissolved oxygen requirements of aquatic life with emphasis on Canadian species: a review. Journal of the
Fisheries Research Board of Canada 32, 2295–2332.
Deutsch, C., Emerson, S., Thompson, L., 2006. Physical-biological interactions in North Pacific oxygen variability. Journal of
Geophysical Research 111. doi:10.1029/2005JC003179.
198
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
Dodimead, A.J., Favorite, F., Hirano, T., 1963. Salmon of the North Pacific Ocean Part II: Review of Oceanography of the subarctic
Pacific Region. International North Pacific Fisheries Commission Bulletin 13, 195.
Emerson, S., Mecking, S., Abell, J., 2001. The biological pump in the subtropical North Pacific Ocean: nutrient sources, Redfield rations
and recent changes. Global Biogeochemical Cycles 15, 535–554.
Emerson, S., Watanabe, Y.W., Ono, T., Mecking, S., 2004. Temporal trends in apparent oxygen utilization in the upper pycnocline of the
North Pacific: 1980–2000. Journal of Oceanography 60, 139–147.
Feely, R.A., Sabine, C.L., Schlitzer, R., Bullister, J.L., Mecking, S., Greeley, D., 2004. Oxygen utilization and organic carbon
remineralization in the upper water column of the Pacific Ocean. Journal of Oceanography 60, 45–52.
Freeland, H.J., 2007. What proportion of the North Pacific Current finds its way into the Gulf of Alaska? Atmosphere-Ocean 44, 321–330.
Freeland, H.J., Cummins, P.F., 2005. Argo: a new tool for environmental monitoring and assessment of the world’s oceans, an example
from the N.E. Pacific. Progress in Oceanography 64, 31–44.
Freeland, H.J., Denman, K.L., 1982. A topographically controlled upwelling center off southern Vancouver Island. Journal of Marine
Research 40, 1069–1093.
Freeland, H., Denman, K., Wong, C.S., Whitney, F., Jacques, R., 1997. Evidence of change in the winter mixed layer in the Northeast
Pacific Ocean. Deep-Sea Research 44, 2117–2129.
Fukasawa, M., Freeland, H., Perkin, R., Watanabe, T., Ichida, H., Nishina, A., 2004. Bottom water warming in the North Pacific Ocean.
Nature 427, 825–827.
Grantham, B.A., Chan, F., Mielsen, K.J., Fox, D.S., Barth, J.A., Huyer, A., Lubchenco, J., Menge, B.A., 2004. Upwelling-driven
nearshore hypoxia signals ecosystem and oceanographic changes in the northeast Pacific. Nature 429, 749–754.
Gray, J.S., Wu, R.S., Or, Y.Y., 2002. Effects of hypoxia and organic enrichment on the coastal marine environment. Marine Ecology
Progress Series 238, 249–279.
Imasoto, N., Kobayashi, T., Fujio, S., 2000. Study of water motion at the dissolved oxygen minimum layer and local oxygen consumption
rate from the Lagrangian viewpoint. Journal of Oceanography 56, 361–377.
Kang, D-J., Kim, J-Y., Lee, T., Kin, K-R., 2004. Will the East/Japan Sea become an anoxic sea in the next century? Marine Chemistry 91,
77–84.
Kumamoto, Y., Murata, A., Watanabe, S., Fukasawa, M., Yoneda, M., Shibata, Y., Morita, M., 2004. Preliminary results of radiocarbon
measurements during the WHP P17N re-visit in 2001. Nuclear Instruments and Methods in Physics Research B, 441–445.
Mackas, D.L., Denman, K.L., Bennett, A.F., 1987. Least squares multiple tracer analysis of water mass composition. Journal of
Geophysical Research 92 (C3), 2907–2918.
Mecking, S., Warner, M.J., Bullister, J.L., 2006. Temporal changes in pCFC-12 ages and AOU along two hydrographic sections in the
eastern subtropical North Pacific. Deep-Sea Research I 53, 169–187.
Ono, T., Midorikawa, T., Watanabe, Y.W., Tadokoro, K., Saino, T., 2001. Temporal increase of phosphate and apparent oxygen
utilization in the subsurface waters of the western subarctic Pacific from 1968 to 1998. Geophysical Research Letters 28, 3285–3288.
Reynolds, R.W., Marsico, D.C., 1993. An improved real-time global sea surface temperature analysis. Journal of Climate 6, 114–119.
Royer, T.C., Grosch, C.E., 2006. Ocean warming and freshening in the northern Gulf of Alaska. Geophysical Research Letters 33,
L16605. doi:10.1029/2006GL026767.
Tabata, S., 1989. Trends and long-term variability of ocean properties at Ocean Station P in the northeast Pacific Ocean. Geophysical
Monographs 55, 113–132.
Tabata, S., Brown, R.M., 1994. Hydrographic/CTD observations made during ocean climate monitoring study: 1981–1991 – a summary
of operational phase of study. Canadian Data Report of Hydrography and Ocean Sciences 136, 136.
Tabata, S., Peart, J.L., 1985. Statistics of oceanographic data based on hydrographic/CTD casts made at Ocean Station P during August
1956 through June 1981. Canadian Data Report of Hydrography and Ocean Sciences 31, 133.
Turner, R.E., Rabalais, N.N., 1994. Coastal eutrophication near the Mississippi river delta. Nature 368, 619–621.
Ueno, H., Yasuda, I., 2003. Intermediate water circulation in the North Pacific subarctic and northern subtropical regions. Journal of
Geophysical Research 108 (C11), 3348. doi:10.1029/2002JC001372.
VanGeen, A., Smethie, W.M., Horneman, A., Lee, H., 2006. Sensitivity of the North Pacific oxygen minimum zone to changes in ocean
circulation: a simple model calibrated by chlorofluorocarbons. Journal of Geophysical Research 111, C10004. doi:10.1029/
2005JC003192.
von Storch, H., Zwiers, F.W., 2002. Statistical Analysis in Climate Research. Cambridge University Press, ISBN 0 521 45071 3, 494.
Watanabe, Y.W., Ono, T., Shimamoto, W., Sugimoto, T., Wakita, M., Watanabe, S., 2001. Probability of a reduction in the formation
rate of the subsurface water in the North Pacific during the 1980s and 1990s. Geophysical Research Letters 28, 3289–3292.
Watanabe, Y.W., Wakita, M., Maeda, N., Ono, T., Gamo, T., 2003. Synchronous bidecadal periodic changes of oxygen, phosphate and
temperature between the Japan Sea deep water and the North Pacific intermediate water. Geophysical Research Letters 30.
doi:10.1029/2003GL018338.
Whitney, F., Barwell-Clarke, J., 1997. WOCE section P15N: hydrographic section of the Pacific Ocean from Dutch Harbor, Alaska, to
American Samoa. Canadian Technical Report of Hydrography and Ocean Sciences 184, 61.
Whitney, F.A., Freeland, H.J., 1999. Variability in upper-ocean water properties in the NE Pacific Ocean. Deep-Sea Research II 46, 2351–
2370.
Whitney, F.A., Robert, M., 2002. Structure of Haida eddies and their transport of nutrient from coastal margins into the NE Pacific
Ocean. Journal of Oceanography 58, 715–723.
Whitney, F.A., Wong, C.S., Boyd, P.W., 1998. Interannual variability in nitrate supply to surface waters of the Northeast Pacific Ocean.
Marine Ecology Progress Series 170, 15–23.
F.A. Whitney et al. / Progress in Oceanography 75 (2007) 179–199
199
Whitney, F.A., Crawford, W.R., Harrison, P.J., 2005. Physical processes that enhance nutrient transport and primary productivity in the
coastal and open ocean of the Subarctic NE Pacific. Deep-Sea Research II 52, 681–706.
Wong, C.S., Matear, R.J., Freeland, H.J., Whitney, F.A., Bychkov, A.S., 1998. WOCE line P1W in the Sea of Okhotsk 2 CFCs and the
formation rate of intermediate water. Journal of Geophysical Research 103, 15625–15642.
Wong, C.S., Whitney, F.A., Crawford, D.W., Iseki, K., Matear, R.J., Johnson, W.K., Page, J.S., Timothy, D., 1999. Seasonal and
interannual variability in particle fluxes of carbon, nitrogen and silicon from time series of sediment traps at Ocean Station P, 19821993: relationship to changes in subarctic primary productivity. Deep-Sea Research II 46, 2735–2760.
Wong, A.P.S., Bindhoff, N.L., Church, J.A., 2001. Freshwater and heat changes in the North and South Pacific Oceans between the 1960s
and 1985-94. Journal of Climate 14, 1613–1633.
Yasuda, I., Osafune, S., Tatebe, H., 2006. Possible explanation linking 18.6-year period nodal tidal cycle with bi-decadal variation of
ocean and climate in the North Pacific. Geophysical Research Letters 33, L08606. doi:10.1029/2005GL025237.
Yoshikawa, C., Nakatsuka, T., Wakatsuchi, M., 2006. Distribution of N* in the Sea of Okhotsk and its use as a biogeochemical tracer of
the Okhotsk Sea Intermediate Water formation process. Journal of Marine Systems 63, 49–62.
You, Y., 2005. Unveiling the mystery of North Pacific Intermediate Water formation. Eos 86 (65), 71.